See other failures under 000.html.
From http://sabrosacycles.blogspot.com/2010/02/aluminum-has-no-fatigue-limit.html as of 2012/08.
Failure noticed failure when standing to lift the front wheel and the pedal and crank end fell off. No injuries.
Rider weight and riding habits unknown.
The owner notes aluminum has no fatigue limit, but that is not typically an issue in bicycle failures. A "fatigue limit" is a stress level such that repeated flexing below that level cannot ever cause a fatigue failure. However, light weight is highly valuable for bicycles, so most parts are designed to operate above the fatigue limit. Even though steel does have a fatigue limit, steel parts regularly fail from fatigue. Thus, the lack of fatigue limit probably not the leading cause of this failure.
The failure location is unusual because it does not occur at a major feature. Failures typically occur at some "feature" or change in section, typically at the pedal eye, spindle eye, spider attachment, or some substantial cross-section change such as the start or end of a weight-reducing (but stress-inducing) groove.
The outside face of the arm is substantially flat to the left of the break and substantially rounded to the right. Further, there are apparent wear marks on the face, probably from repeated occasinal heel rubbing on the arm. These likely contributed to the failure.
The section photo of the crack shows darkening on a diagonal line towards the upper center of the photo. Darkening indicates where the crack already was for a while, while the light part shows the last part to fail.
When the rider's foot is near the bottom of the stroke, there are often large bending forces along the crank, carried mostly by the outside face. A smaller the outside face means the load is carried by less metal, thus increasing failure rates. It is possible the failure happened because (a) the original design has a rounded face that concentrates the load; (b) heel rub caused a section change that increased local stresses; (c) an initial microsopic crack was "grown" by a combination of forces including mid-stroke forces that caused the crack to grow towards the upper edge.
If that is the cause of failure (if, not a sure thing), then the rider would have gotten durability from a crank with a more rectangular cross-section, even if doing so slightly reduced one or both dimensions (and thus made the crank slightly more flexible).
See also FAIL-170.html